[parser] LPG v2 into Orbit

[eclipse-ocl-bot]

| --- | --- | | Bugzilla Link | 288666 | | Status | CLOSED FIXED | | Importance | P3 enhancement | | Reported | Sep 04, 2009 13:01 EDT | | Modified | May 27, 2011 02:47 EDT | | Version | 1.3.0 | | Blocks | 242153 | | Reporter | Adolfo Sanchez-Barbudo Herrera |

Description

Make LPGv2 take part in Orbit so that new OCL grammars, parsing infrastructure and textual editor based on this LPG runtime may be integrated in MDT-OCL project.

In principle, LPG 2.0.18 will be our target.

[eclipse-ocl-bot]

By Adolfo Sanchez-Barbudo Herrera on Dec 10, 2009 08:02

Since LPG 2.0.18 release has been delayed, our target will be 2.0.17.

[eclipse-ocl-bot]

By Ed Willink on Dec 12, 2009 07:04

Bug 297659 has been marked as a duplicate of this bug.

[eclipse-ocl-bot]

By Adolfo Sanchez-Barbudo Herrera on Dec 22, 2009 11:07

CQ approved: http://dev.eclipse.org/ipzilla/show_bug.cgi?id=3653

Next step: Check in the new LPG Runtime into Orbit's repository

As agreed with Robert's Fuhrer, the bundle's name will be: lpg.runtime.java

[eclipse-ocl-bot]

By Adolfo Sanchez-Barbudo Herrera on Dec 22, 2009 11:47

Created attachment 154938 LPG Runtime inclusion patch

The attached patch should be applied against

dev.eclipse.org/tools/org.eclipse.orbit/lpg.runtime.java project BRANCH v2_0_17

The bundle has been created in a similar way the previous net.sourceforge.lpg.lpgjavaruntime bundle was created.

The following guidelines have also been taken:

http://wiki.eclipse.org/Adding_Bundles_to_Orbit

:notepad_spiral: ocl288666_LPGintoOrbit.patch

[eclipse-ocl-bot]

By Adolfo Sanchez-Barbudo Herrera on Dec 22, 2009 12:14

Adding David...

David, since it's my first contribution to Orbit I would be pleased if you are able to do a quick review of the contribution.

I think that I'll need tag once this is committed to properly update the bundles.map

A patch with iplog, bundles.map, orbit.psf and set1/features.xml will be provided when this one is committed.

[eclipse-ocl-bot]

By Adolfo Sanchez-Barbudo Herrera on Dec 22, 2009 12:17

Also adding Robert since he must be interested in this.

[eclipse-ocl-bot]

By David Williams on Dec 22, 2009 13:11

Looks close enough to try. I didn't see any obvious errors or common mistakes (with a 4 minute review).

Don't forget, you'll also have to add it to the feature.xml file and map file too.

I'll be "around" most the time, so I think its fine if you just give it a try, and if there's a few build failures, this is a good time to have them (not a critical time of milestone) and I'm sure you'll straighten it out ... I mean, I'll be here and happy to help if you need it. :)

Thanks.

[eclipse-ocl-bot]

By Adolfo Sanchez-Barbudo Herrera on Dec 22, 2009 14:01

David,

Thank you very much !!

I think I'm done. How could I check that all has gone fine ?. I remember reading that I have to wait 1 hour so that Orbit's realizes that it has a new bundle to build. Am I wrong ?

Cheers,\ Adolfo.

[eclipse-ocl-bot]

By Mat Booth on Dec 22, 2009 14:08

Adolfo,

For my own edification, would you be able to outline the chief benefits in moving to LPG2?

Regards,\ Mat

[eclipse-ocl-bot]

By David Williams on Dec 22, 2009 14:25

(In reply to comment #8)

... I remember reading that I have to wait 1 hour so that Orbit's realizes that it has a new bundle to build. Am I wrong ?

Yes, up to one hour. But if you are impatient, you can "force a build" by clicking on the refreshing-looking button at \ http://build.eclipse.org:9777/dashboard/tab/builds

But, one has started, and will be done in hour or two.

[eclipse-ocl-bot]

By David Williams on Dec 22, 2009 14:26

Oh, and if the build works at all, you can see/get the results from \ http://download.eclipse.org/tools/orbit/committers/

[eclipse-ocl-bot]

By David Williams on Dec 22, 2009 14:29

You'll want to be sure this runs as expected. When I pulled in the code from cvs, I later noticed a bunch of messages in my log file. All of this form:

!ENTRY org.eclipse.jdt.core 2 0 2009-12-22 13:08:25.556\ !MESSAGE The Java indexing could not index /lpg.runtime.java/lpg/runtime/Recover\ yParser.class. This .class file doesn't follow the class file format specificati\ on. Please report this issue against the .class file vendor

So ... think these are really odd .class files? Could you have accidental added end-of-lines in them somehow?

[eclipse-ocl-bot]

By Adolfo Sanchez-Barbudo Herrera on Dec 22, 2009 14:36

(In reply to comment #12)

You'll want to be sure this runs as expected. When I pulled in the code from cvs, I later noticed a bunch of messages in my log file. All of this form:

!ENTRY org.eclipse.jdt.core 2 0 2009-12-22 13:08:25.556 !MESSAGE The Java indexing could not index /lpg.runtime.java/lpg/runtime/Recover yParser.class. This .class file doesn't follow the class file format specificati on. Please report this issue against the .class file vendor

So ... think these are really odd .class files? Could you have accidental added end-of-lines in them somehow?

I think I have simply added the classes (and the .java files in the source-bundle), from the lpg.runtime.jar downloaded from sourceforge.net... I'll check it again to ensure they are same...

[eclipse-ocl-bot]

By Adolfo Sanchez-Barbudo Herrera on Dec 22, 2009 14:42

I think I have simply added the classes (and the .java files in the source-bundle), from the lpg.runtime.jar downloaded from sourceforge.net... I'll check it again to ensure they are same...

I'll try to recompile the .java file with my Java 5 compiler, and check if they are the same.

Cheers,\ Adolfo.

[eclipse-ocl-bot]

By Adolfo Sanchez-Barbudo Herrera on Dec 22, 2009 14:57

I'll try to recompile the .java file with my Java 5 compiler, and check if they are the same.

Cheers, Adolfo.

After:

1. recompiling the source-bundle .java files in a temporal bin folder.
2. replacing the original root-project .class files by the new .class files.
3. Comparing the project the Orbit's cvs.

Only a non-perceptible-by-the-compare-editor change is detected for the messages.properties file. Do you think this could be error's source ?

[eclipse-ocl-bot]

By David Williams on Dec 22, 2009 15:40

(In reply to comment #15)

... Do you think this could be error's source ?

No, I wouldn't think so. I'm using eclipse SDK M4 as my development environment. I wonder if this is some new JDT check? Maybe their bug? Maybe would happen to lots of .class files? (not compiled by JDT :)

If you haven't done something funny (like use FTP to get the .class files, or change your .class cvs filetype setting to ascii (both, highly unlikely ... that's all I can think of of how you'd cause this) then it may not be a real problem.

If someone else sees (after using eclipse-sdk M4) perhaps we could ask someone on JDT team "what it means" and if we should be concerned. As far as I know, it could just be my setup, somehow.

[eclipse-ocl-bot]

By David Williams on Dec 22, 2009 15:46

(In reply to comment #16)

(In reply to comment #15) \

... Do you think this could be error's source ?

... it could just be my setup, somehow.

Oh wait, I think I know what caused this. I initially didn't check out from cvs, but just applied your patch. And "binary" files don't actually patch correctly, or at least not easily ... so probably did have some extra EOLs in there, or something. I imagine you actually "imported" these into your workspace and then committed, so should be fine.

I do not see this message now that I've checked out your version directly from cvs.

Sorry for the false alarm.

[eclipse-ocl-bot]

By Ed Willink on Dec 22, 2009 17:05

For my own edification, would you be able to outline the chief benefits in moving to LPG2?

a) responding from a plea from the LPG developers to stop using LPG 1\ b) using the active development stream\ c) using the same stream as used by IMP

Of the technical changes, the only one that I know of that seems useful for MDT/OCL's mundane purposes is:

d) the possibility to use type-safe grammar rule returns

[eclipse-ocl-bot]

By Adolfo Sanchez-Barbudo Herrera on Dec 23, 2009 04:34

Oh wait, I think I know what caused this. I initially didn't check out from cvs, but just applied your patch. And "binary" files don't actually patch correctly, or at least not easily ... so probably did have some extra EOLs in there, or something. I imagine you actually "imported" these into your workspace and then committed, so should be fine.

I do not see this message now that I've checked out your version directly from cvs.

Sorry for the false alarm.

It doesn't matter. It seems that all has gone fine. I have checked that the binary and source bundles have been sucessfully built.

Thank you very much for the help and congratulations for the helpful documentation, it has been essential to make this work in my first try.

N.B: I'll wait until LPGv2.0.17 reuse approval for MDT-OCL, to resolve the bug.

[eclipse-ocl-bot]

By Adolfo Sanchez-Barbudo Herrera on Dec 23, 2009 04:58

(In reply to comment #18)

For my own edification, would you be able to outline the chief benefits in moving to LPG2?

a) responding from a plea from the LPG developers to stop using LPG 1 b) using the active development stream c) using the same stream as used by IMP

Of the technical changes, the only one that I know of that seems useful for MDT/OCL's mundane purposes is:

d) the possibility to use type-safe grammar rule returns

There is a long thread in which you may find more reasons. If you are bored you could try to follow Bug 242153. Apart from all that Ed has pointed out, I could add some notes:

• We tend to ship an OCLEditor as an example which uses IMP, which uses LPGv2. So, since LPGv2 will be needed by the example it should be good that the parser uses the same runtime, and we could then forget the old runtime.
• Technically, LPGv2.x runtime has been refactored improving its design, including more interfaces, etc which should also be exploited by MDT-OCL. At least, we will do the first step, changing the runtime in this major version increment (MDT-OCL 3.0.0). Let's see if we are able to do notable changes to make MDT-OCL parser take advantage from the new LPG design.
• Technically, LPGv1.x runtime has some deficiences relating performance and memory usage. I detected one related to backtracking parser, which used an IntTuple action = new IntTuple(1 << 20) to stack parser actions. In Lpg v2.x an improved dinamic array is used instead, which shouldn't use so amount of memory: private IntSegmentedTuple action = new IntSegmentedTuple(10, 1024);

Regards,\ Adolfo.

[eclipse-ocl-bot]

By Adolfo Sanchez-Barbudo Herrera on Dec 24, 2009 04:59

Approval granted : http://dev.eclipse.org/ipzilla/show_bug.cgi?id=3656

Fixed

[eclipse-ocl-bot]

By Ed Willink on May 27, 2011 02:47

Closing after over 18 months in resolved state.